Paying With Money or With Effort: Pricing When Customers Anticipate Hassle Anja Lambrecht* London Business School
[email protected]
Catherine Tucker+ MIT Sloan School of Management
[email protected] *Anja Lambrecht, Assistant Professor of Marketing, London Business School, Regent’s Park, London NW1 4SA, United Kingdom, tel: +44 (0) 20 7000 8645, fax: +44 (0) 20 7000 8601, email:
[email protected] +
Catherine Tucker, Assistant Professor of Marketing, MIT Sloan School of Management, 100 Main St E62-533, Cambridge MA 02142, USA, tel: +1 (617) 252 1499, fax: +1 (617) 258-7597, email:
[email protected]
Anja Lambrecht thanks Deutsche Forschungsgemeinschaft, the London Business School Center of Marketing and the Institute of Innovation and Entrepeneurship. The paper benefited from discussions with Marco Bertini, Simona Botti, Rajesh Chandy, David Faro, Shane Frederick, Avi Goldfarb, Monika Heller, Dan Goldstein, Kanishka Misra, John Roberts, Nader Tavassoli and Juanjuan Zhang. Keywords: Pricing, Services, Field Experiment, Contractual Discounting, Hassle Costs, PeriodLevel Bracketing
Electronic copy available at: http://ssrn.com/abstract=1805109
2
Abstract For many services, customers subscribe to long-term contracts. Standard economic theory suggests that customers evaluate a contract as the sum of benefits and payments. We suggest that rather than evaluating multi-period service contracts at the contract-level, customers use periodlevel bracketing. They evaluate the distinct per-period loss or gain they incur from choosing this contract. This has important consequences when benefits vary over the course of the contract, for example due to “hassle costs.” If customers use period-level bracketing, they will value a lower price more in periods where they have hassle than in other periods. We explore this using data from a field experiment for web hosting services. The field experiment had 2 hassle cost priming conditions (present, absent) x 2 discount conditions (offered, not offered). We find that a lower price in the initial period is more attractive to customers when they expect their hassle costs to be high at setup. In five lab experiments, we support and extend the field experiment’s findings. We find evidence for period-level bracketing when customers have hassle costs, independently of whether hassle costs occur in the first, an intermediate or the last period of a contract. We rule out alternative explanations, such as hyperbolic discounting. Our findings suggest that in setting prices, firms should consider the timing of hassle costs faced by customers.
Electronic copy available at: http://ssrn.com/abstract=1805109
3
INTRODUCTION For many services, such as cell phone, Internet, television or web services, customers commit to a long-term contract with a firm that spans multiple time periods, typically several months. Standard economic theory assumes that a customer evaluates payments and benefits at the contract level and is indifferent about the distribution of payments and benefits within a contract, so long as the total net discounted value of the contract remains the same. We suggest that in multi-period service contracts, customers instead evaluate benefits and payments using period-level bracketing. They evaluate the distinct per-period loss or gain they incur from choosing this contract. As a result, under period-level bracketing customers are not indifferent to the distribution of payments and benefits over time. Period-level bracketing would affect decision-making when the value of a service varies over time. An important source of such variation is hassle costs. In service contracts, customers often experience hassle in distinct periods. Hassle can arise when setting up a service, such as installing a wireless network at home; maintaining an ongoing service, such as waiting for the satellite TV company to visit and fix a broken connection; or at the end of the contract, like when cancelling a telephone contract. We define “hassle costs” as the non-monetary effort and inconvenience a customer incurs in setting up, maintaining or disposing of a product or service. Hassle costs may prevent customers from entering into contracts or reduce their satisfaction with their existing provider. If customers use period-level bracketing, they will care when hassle occurs within a contract, and should prefer lower payments, for example discounts, at the time of higher hassle costs. This contrasts with standard economic theory that would predict that preferences for a payment schedule should be independent from when customers face hassle costs.
4 In our field experiment, we use search advertising to manipulate and consequently
identify the effect of customer expectations of hassle costs on pricing plan choice. In five subsequent lab experiments, we vary the timing of hassle costs and the offered pricing plans. Our results provide strong evidence that customers use period-level bracketing, preferring to match the pattern of their payments to the pattern of their hassle costs whether hassle costs arise at the beginning, during or at the end of their service contract. This reflects the fact that firms may impose hassle costs on their consumers during the setup, maintenance, or disposal of a product. They also illustrate how firms can benefit from designing pricing schemes that reduce the payment that a customer makes in the period of hassle, even if the overall amount the firm charges over the contract remains the same.
CONCEPTUAL FRAMEWORK Decision-Making for Contracts Standard economic principles suggest that customers evaluate a contract based on the net present value of the sum of benefits and payments within a contract. Without loss of generality, let us assume that the discount rate is zero. In period t a customer receives benefit xt from a service, which provides a level of utility, g ( xt ) , and pays the price pt . The coefficient denotes the transformation of monetary units into utility. The net value of a contract is then evaluated as (1) VTC g ( xt ) pt . t
We refer to this as contract-level bracketing since the evaluation is based on the total outcome of the contract. We argue that instead consumers may evaluate contracts using periodlevel bracketing. Previous research suggests that consumers tend to evaluate outcomes in distinct time periods, often in relatively narrow time frames (Benartzi and Thaler 1999; Read et al.
5
1999). This is true even if events are part of a decision or strategy that spans multiple time periods. For example, there is evidence that individuals’ evaluation period of stock market returns are substantially shorter than the full span of their investment horizon (Benartzi and Thaler 1995). Within these evaluation periods, customers aim to achieve the best possible outcomes. These period-level outcomes guide overall decisions. Therefore when customers are faced with a contract, they evaluate its attractiveness in individual time periods, for example associated with the customary billing cycle, instead of evaluating the full contract as a whole. In each separate period, they wish to achieve the best possible outcome rather than demanding a positive outcome only over the total course of the contract. Prospect theory suggests that rather than evaluating the absolute outcome of a contract, consumers evaluate outcomes as a gain or a loss relative to a reference point (Kahneman and Tversky 1979). If v . represents the value a consumer assigns to the per-period outcome of a contract, the value of this contract across all periods under contract-level bracketing is: (2) VTP v g ( xt ) pt . t
A negative per-period outcome is then coded as a loss, a positive outcome as a gain. Period-level bracketing relates to past research that provides evidence that customers often do not evaluate the joint outcomes of all choices or events (Kahneman and Tversky 1979; Thaler 1985), but instead bracket them narrowly, separately evaluating the outcomes of multiple choices (Tversky and Kahneman 1981; Read et al. 1999; Thaler 1999). Our work differs in that we extend the concept of bracketing to how consumers evaluate the decision on a stream of predetermined benefits and payments across distinct time periods. We examine whether they evaluate this decision based on its aggregate outcome or based on the outcomes of individual brackets, typically different time periods.
6
Hassle Costs The benefit customers receive from a service may vary across time periods, often due to hassle costs. Customers usually experience hassle in distinct periods, such as to set up a service, to maintain a service, or to dispose of it at the end of the contract. We define hassle costs as the non-monetary effort and inconvenience a customer incurs in setting up, maintaining or disposing of a product or service. Firms should manage hassle costs actively for three reasons. First, hassle costs are disliked by and salient to customers. Second, the firm can usually predict and manage the timing of hassle its customers’ experience. Third, the firm can manage the level of hassle or, as we study in this paper, alleviate its consequences by adjusting its pricing schemes.
The Effect of Hassle Costs on Decision-Making Period-level and contract-level bracketing lead to different choices when consumers have hassle costs. For ease of exposition, assume a two-period model and a discount rate of zero. All benefits from the product, xt , and payments, pt , are the same in periods 1 and 2. Let h1 be the hassle costs in period 1. Assume the firm holds the total bill of the contract, ( p1 p2 ), constant but is indifferent to the distribution of payments over time. It can vary the price pt by the amount t . This means that the firm can discount the price in the period of hassle, p1 , by 1 and increase the price in the other period, p2 , by 2 . It sets 1 2 . If customers use contractlevel bracketing, they are indifferent to how the firm distributes payments over time with respect to hassle, since a change to the flow of payments would not affect the overall contract’s value: (3) VT ,d ,h g ( x1 h1 ) ( p1 1 ) g ( x2 ) ( p2 2 ) g ( x1 h1 ) p1 g ( x2 ) p2 C
This is, however, different under period-level bracketing. If monthly payments do not vary, i.e. 1 2 0 , consumers would code the outcome of period 1 as a lower gain than the
7
outcome of period 2, or as a loss. By decreasing the price in period 1 by 1 and increasing the price in period 2 by 2 , as illustrated in equation (4), the total value to the consumer increases since the marginal value per time period is lower for greater gains than for smaller gains. (4) VT ,d ,h v g ( x1 h1 ) ( p1 1 ) v g ( x2 ) ( p2 2 ) . P
Therefore if there are hassle costs, the value of a contract that discounts prices in the period of hassle and charges more in the period of no hassle is higher than that of a contract with even prices in all periods. Figure 1(A) further illustrates this: The increase in period 1 value from lower payment 1 , A, is greater than the decrease in period 2 value from greater payment 2 , B. If the outcome of period 1 represents a loss rather than a lower gain, the effect of shifting payments across periods is even more pronounced (see Figure 1(B): The increase in period 1 value from lower payment 1 , A, is greater than the decrease in period 2 value from greater payment 2 , B.) Consequently, customers who use period-level bracketing and face hassle costs prefer contracts that match periods of lower overall benefits with lower payments. This allows them to have a pattern of monetary costs that reflects the pattern of their hassle costs on a per-period level. This holds if hassle costs and lower payments occur in the second instead of the first period (in Figure 1 (A) and (B), simply reverse periods 1 and 2). Similarly, if the firm charges a fixed per-period price but discounts any one period of the contract, customers with period-level bracketing always prefer to obtain the discount in the period of hassle. Last, if customers’ discount rates are positive, the preference for an early-period discount increases and the preference for a later discount decreases, but our general result still holds. Our discussion focuses on hassle costs because of their unique importance to firms. However, shocks other than hassle costs to a customer’s experience utility may have a similar
8
effect. Period-level bracketing may then explain choice in other contractual settings. For example, period-level bracketing may explain why customers prefer to match the duration of a loan with the lifetime of the product that the loan finances (Hirst et al. 1994). More broadly, it can explain why New York City cabdrivers make labor decisions “one day at a time” instead of intertemporally substituting labor and leisure across multiple days (Camerer et al. 1997). We refer to the strategy of granting lower payments in one period and recapturing the amount in other periods as contractual discounting. Customers know all prices ex ante and commit to the full schedule of prices, so a firm can use this pricing strategy to respond to customers’ hassle costs. Even when the price discount occurs in the first period, contractual discounting differs from typical introductory pricing discussed by the switching costs literature (Klemperer 1995), because higher prices in later periods are contractually mandated and certain.
Relationship to Prior Literature We next clarify how our research adds to four different but related streams of literature. First, our conceptual model relates to research on how customers react to payments and benefits that occur at different points in time (Prelec and Loewenstein 1998). This work focuses on the reciprocal interactions between the benefits (“pleasure”) and costs (“pain”) of a choice. It shows that the thought of future payment may lessen a customer’s pleasure. In contrast to our work, it assumes contract-level, not period-level bracketing. As a result, customers would be indifferent between matching hassle costs with lower per-period prices and constant pricing. Second, we contribute period-level bracketing as one motivation for why firms offer contractual discounting. It is not our objective to analyze more broadly why firms offer or how customers choose between constant pricing, early discounts or late discounts. Customers’ time preferences may make them attach higher value to lower payments in early periods (Olson and
9
Bailey 1981; Loewenstein 1987; Frederick et al. 2002). The promotional pull from heavily discounting one period may be larger than the pull from slightly discounting all periods (Shampanier et al. 2007). Alternatively, customers who prefer to pay the same bill every period may choose constant pricing over contractual discounting (Lambrecht and Skiera 2006). The preference for contractual discounting may also vary with whether the discount is offered early or late in a contract. The focus of our experimental design is to identify the effect of hassle costs separately from the effect of other factors. Third, by suggesting that firms should vary prices within contracts to reflect customers’ hassle costs, we introduce a new dimension into the analysis of nonlinear pricing policies. This literature traditionally focuses on how firms can use linear, two-part or three-part tariffs to implement second-degree price discrimination (Wilson 1993; Winer 2005). Two-part tariffs allow firms to price-discriminate based on customers’ average usage (e.g., Danaher 2002; Narayanan et al. 2007), whereas three-part tariffs also discriminate along customers’ usage variation (Lambrecht et al. 2007). This literature does not discuss how firms should react to hassle costs, or how period-level bracketing may affect the evaluation of long-term service contracts. Similarly, research on nonlinear pricing, the literature on dynamic prices for new durable products examines situations where prices fall over time as a response to price discrimination along customers’ willingness to pay due to their product valuation, rather than to hassle costs (Nair 2008). Fourth, our research builds on the literature on switching costs. Hassle costs and nonmonetary switching costs can both result from similar sources, such as effort invested in learning how to use a new product that would have to be replicated if the user switched to a new provider. The important distinction is that hassle costs are often independent from the intention to switch.
10
They include, for example, the costs that customers undergo during regular maintenance of a service. By contrast, switching costs only include costs that customers anticipate they will have to incur if they switch providers. The type of behavior we study is also distinct from the switching costs literature. We focus on the effect of hassle costs on customers’ preferences for a sequence of prices which are set in stone by a contract. There is no uncertainty over future prices or events. By contrast, the switching costs literature investigates optimal pricing strategies if customers have monetary or non-monetary costs that make it costly to switch between sellers and if the customers anticipate that they will make further purchases at unspecified prices from the same seller that are complementary to the original purchase (Klemperer 1995; Farrell and Klemperer 2007). Empirical studies have documented that as a result of these switching costs, firms have to compensate customers with lower prices initially because customers anticipate higher prices in the future (Chen and Hitt 200; Viard 2007). Our contribution is to demonstrate that customers prefer lower prices in the period of hassle costs, which may be any period of the contract, even if they do not anticipate higher future prices. This affects their preference for the distribution of payments over time even when the total amount is pre-determined. Our unique insight is that firms have to compensate customers with lower prices in periods of hassle costs because customers engage in period-level bracketing.
FIELD EXPERIMENT Empirical Setting We use data from a field experiment conducted by a web hosting provider in the UK. Web hosting providers sell web space and bandwidth for a monthly fee, allowing their customers to set up and maintain websites (see “Industry Setting” in the Online Appendix). Web hosting
11
providers typically offer contracts that span multiple months. Pricing schemes vary from provider to provider, so there are few customer expectations about pricing schemes. The web-hosting industry is characterized by high uncertainty about hassle costs, in particular about how easy it will be to set up a website. In later periods, customers typically have low or zero hassle costs. Web hosting firms are increasingly turning to search engines to acquire customers. When a customer uses a specific search term, the search engine displays the ad and charges the firm per click from the search engine’s to the firm’s site.
Design, Procedure and Results Design and procedure. This web hosting firm sells various web hosting packages to personal users and small businesses alongside domain names and offers both contractual discounting and constant monthly pricing for contracts that last three months. The firm tested which form of search advertising wording on Google was more effective in attracting serious customers. This field experiment tests how prospects react to the promise of a lack of hassle compared to an otherwise expected level of hassle; later lab experiments will test period-level bracketing when consumers expect a high level of hassle. We expect that firms’ promise of initial discounting is more attractive to customers when firms do not explicitly communicate a lack of hassle. Figure 2 displays the four advertising messages used. In the first line of the ad, the firm gave either no explicit information on the hassle of setup (“Web Host”) or explicitly communicated "no hassle" (“Web Host, No Hassle Setup”). The second line indicated the price. Ads 1 and 3 offered a monthly constant price of £1.29. Ads 2 and 4 offered contractual discounting where the first month levies a charge of £0.00 and the following months a monthly price of £1.99. Prices were set so that cumulatively, over three months, the total bill under contractual discounting would be £0.11, or 2.8%, higher than
12
under constant monthly pricing. We show robustness to differences in the total bill amount in the subsequent lab experiments. In the third line of the ad, “3 mo. contracts” indicates that a contract lasts for three months. “Instant setup” indicates that webspace is allocated automatically, so that once the customer has set the website up it appears instantaneously on the Internet (see “Testing the Advertising Messages” in the Online Appendix). Customers who clicked on any Google ad were directed to the same website of the provider, which offered all advertised hosting plans. The four ads were tested on Google in the UK over the course of seven days. Google rotated the ads automatically so that they were displayed an equal number of times. The ad messages were associated with 14 different search terms (Table 1). There was large variation in both the number of impressions (the number of web searchers who saw the ad because they used that search term), and the click-through rates (the number of web searchers who, having used that search term, clicked on the ad). Over 79,000 web searchers were exposed to the different randomized treatments. We checked that there was no difference in how technical the search terms were that users used to reach the website by condition. There was no statistical difference between the two hassle-conditions (p-value=0.87) or between the two pricing-conditions (p-value=0.15). Our conceptual framework predicts that, for a customer who anticipates low hassle cost, the pattern of prices with initial discounts does not mirror the pattern of hassle costs. Such a customer should be less likely to be attracted by a contract that offers contractual discounting than a customer who views an ad that does not provide information about hassle costs. We therefore expect ads that advertise both separately to perform significantly better than the ad that jointly advertises “no hassle” and contractual discounting or the ad that advertises neither.
13 We evaluate the empirical data on three levels. First, we analyze how likely customers that
viewed the different ads were to click through to the firm’s website. Since Google does not provide information on the success of each ad variation by keyword, but only in aggregate, we are not able to include keywords as controls in our click-through analysis. Second, we analyze subsequent browsing behavior once customers arrived on the website. Third, we stratify our analysis of browsing behavior by what search term people used to reach the website.
Results based on click-through rates. Table 2 displays the click-through rates (CTR) for each of the ad variations from Figure 2. The least successful ad was the one that combined both a message of "no hassle" and contractual discounting (CTR 0.40%). In line with our expectation, the raw mean CTR suggest that the most successful ads were the ads that offered either contractual discounting (CTR 0.53%) or "no hassle" (CTR 0.53%). Table 3 displays a probit specification that compares the performance of the ads in order to evaluate the statistical differences between the rates displayed in Table 3. The dependent variable is whether a visitor to Google who saw the ad clicked through on the ad. The base condition is the worst-performing ad ("no hassle" and contractual discounting). As expected, the analysis shows that relative to the baseline ad that offers both "no hassle" and contractual discounting, it was more effective to either advertise a lack of hassle or to advertise contractual discounting in isolation (p
